This document explores the different ways in which a grid of points covering Antarctic ice-free areas can be classified into different habitat types.

The variables available for analysis are coast, geoT, cloud, sumtemp, wind, temp, melt, modT_0315, elev, rad, rugos, slope, DDminus5, precip

Habitats were classified using a fuzzy clustering algorithm, cmeans. We classified habitats using all variables and two different subsets of the variables.

In the plots below. The title of each legend can be read as follows: The first number refers to the number of clusters set into the cmeans algorithm. W and U refer to weighted and unweighted cmeans. The last number refers to which subset of variables was used. The first subset of variables is: elevation, rugosity, precipitation, temperature, distance to coast, and wind. The second subset of variables is: distance to coast, distance to geothermal site, cloud cover, wind, temperature, elevation, slope, and precipitation. Lack of a number means all variables were used.

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

After meeting on 15 June 2020, we decided to redo habitat classifications with the following variable set:

Elevation [“elev”], Rugosity [“rugos”], precipitation [“precip”], degree days [“DDminus5”], distance to coast [“coast”], wind [“wind”], and solar radiation [“rad”].

Resulting plot below:

vars <- c("elev", "rugos", "precip", "DDminus5", "coast", "wind", "rad")

x <- princomp(envpred_norm[,2:15] %>% select(vars))

autoplot(x, data = habDat, col = "n8W3", 
         loadings = TRUE, loadings.label = TRUE, 
         loadings.label.size = 5, alpha = 0.4)